Acidic transcription factors position the genome at nuclear speckles through transcription dependent and independent mechanisms

SUMMARY

A small fraction of the genome positions reproducibly near nuclear speckles (NS), increasing expression and/or splicing efficiency of NS-associated genes. How specific genomic regions in mammalian cells target to NS remains unclear. Here we demonstrate establishment of genome-wide NS-association without active transcription. We show that DNA sequences derived from NS-associated regions integrated as transgenes autonomously target to NS. By systematically dissecting one such genomic locus, the COL1A1-SGCA locus, we identified redundant NS-targeting cis-regulatory elements, including a ~600 bp fragment with 17 binding motifs for 8 transcription factors (TFs). Four NS-targeting TFs within this fragment contain acidic activation domains (AADs) that provide both chromatin-context and transcription-dependent NS-targeting, a property that appears common among several other tested AADs. A subset of acidic activator TFs contain an additional, transcription-independent NS-targeting activity. Our findings establish diverse and partially redundant NS-targeting activities, which may facilitate dynamic gene positioning at NS periphery for context-specific transcriptional responses.

eTOC BLURB

Chaturvedi et al. demonstrate that acidic transcription factors reposition genomic loci to nuclear speckles through both transcription-dependent and transcription-independent mechanisms, revealing that speckle association is actively established by cis-regulatory DNA elements and transcription factor properties rather than being established simply as a consequence of active gene expression.

Graphical Abstract

INTRODUCTION

Nuclear speckles (NS) are membraneless bodies within the metazoan nucleus that concentrate many factors involved in RNA Pol II gene expression and influence their distribution, dynamics, and interactions. These enriched factors participate in multiple stages of RNA Pol II gene expression, including different steps of transcription, RNA processing, and RNA export.^1–5 NS are irregularly shaped with variable size (~0.3-3.0 μm) and numbers (~20-50) per nucleus.^6–9 Depending on detection methods, NS have also been termed hyaline grumes,¹⁰ interchromatin granule clusters (IGCs),^7,11 poly(A) RNA domains,¹² SC-35 domains,⁶ splicing factor compartments (SFCs),¹³ or B snurposomes in amphibian oocytes.¹¹ Even the definition of the NS periphery varies with the NS markers used for their detection.⁵ SON and SRRM2 are considered to be core NS proteins.^14,15 Here we operationally define NS using SON immunostaining, which closely corresponds to the light microscopy equivalent of IGCs.

In situ hybridization of >25 actively transcribed genes previously demonstrated close NS-association of more than half of these genes.^8,14,16 These genes were subdivided into Type 1 genes, whose nascent transcripts accumulate within NS (e.g., COL1A1 and cardiac myosin heavy chain (cMyHC)), and Type 2 genes, whose transcripts accumulate adjacent to the gene without entering NS (e.g., β-actin (ACTB), E2F4, lamin A/C (LMNA), and fibronectin (FN1)).^16,17 These observations led to a “gene expression hub” model, proposing that NS constitute a specialized nuclear compartment facilitating expression of a subset of genes.¹⁴

More recently, genome-wide sequencing (e.g., TSA-Seq and SPRITE)^18–20 and imaging^21,22 approaches confirm preferential positioning of a subset of highly active chromosomal regions near NS. More specifically, TSA-Seq demonstrated approximately half of the top 5% most highly expressed genes are positioned within several hundred nanometers of the NS periphery,¹⁸ a distribution consistent across multiple cell lines.²⁰ Additionally, proximity to NS has been shown to increase gene expression through amplified expression^5,23,24 and/or through enhanced RNA splicing efficiency.^25,26

Disruption of NS by combined SON and SRRM2 depletion leads to downregulation of 100s of NS-proximal genes,²⁷ indicating that NS proximity contributes to their expression. However, the mechanisms directing these loci to NS remain largely unknown. Previously, we demonstrated that HSPA1 BAC transgenes autonomously target to NS,²⁸ similarly to the targeting of the endogenous HSPA1 locus, and this NS-targeting is transcription-dependent.²⁸ In contrast, NS-association of the COL1A1 gene is transcription-independent.¹⁷ Together, these observations suggest the existence of at least two distinct NS-targeting mechanisms.

Here we applied a top-down strategy to identify cis and trans determinants of transcription-dependent or transcription-independent NS-targeting. Using TSA-Seq, we first showed that the establishment and maintenance of chromosome positioning relative to NS is transcription-independent. We then demonstrated that 100-200 kbp DNA segments derived from speckle-associated domains (SPADs) autonomously target to NS after random genomic integration. Using this autonomous targeting activity as an assay, we identified multiple functionally redundant cis regulatory elements (CREs) that mediate either transcription-dependent or -independent NS-targeting. This dissection further revealed that transcription-dependent NS targeting is a common property of acidic activation domains (AADs) within acidic transcription factors (TFs), with a subset of acidic TFs containing other protein domains that confer additional, transcription-independent NS-targeting.

Together, our findings support a model in which the combinatorial action of multiple, functionally-redundant CREs containing acidic TF binding motifs drive the targeting and stable anchoring of specific chromosomal regions to the NS periphery.

RESULTS

Genome organization relative to NS is maintained and established independent of transcription

To test whether transcription is required for establishing and/or maintenance of genome organization relative to NS, we inhibited transcription using Triptolide (TPL) or 5,6-Dichloro-1-β-D-ribofuranosylbenzimidazole (DRB). Both treatments strongly reduced nascent transcription, measured by 5-Ethynyl uridine (EU) incorporation and RNA Pol II Ser5 phosphorylation in asynchronous RPE1 cells (Figure 1A–C). As reported previously,^3,29 inhibitor treatment induced NS rounding, and ~40% of cells after 4-hour DRB treatment showed enlarged NS (Figure 1A).

Figure 1. Genome organization relative to NS is maintained and established independent of transcription.

(A-C) Validation and quantification of transcription inhibition in RPE1 cells treated with DMSO, TPL, or DRB. Normalized EU (B) and Pol II Ser5p (C) nuclear intensities compared to the DRB mean and represented as violin plots (n = 20-23)

(D) Genome browser view of SON TSA-Seq enrichment scores in untreated, DMSO-, TPL-, or DRB- treated RPE1 cells.

(E) Schematic of RPE1 synchronization and transcription inhibition.

(F-G) Validation and quantification of transcription inhibition beginning in mitosis on G1 phase RPE1 cells treated with DMSO or TPL. Normalized nuclear Pol II Ser5P intensities compared to the TPL mean and represented as violin plots (n = 14).

(H-M) 2D histograms comparing SON TSA-Seq scores across replicates, treatments, and cell-cycle states. Colors show 20 kb bin counts (~0.02 × 0.02), Pearson’s r correlation.

(N) SON TSA-Seq score percentile tracks (20 kb bins, top), and segmented Speckle-Associated Domains (SPADs, bottom) in log phase or G1 phase RPE1 cells.

(O–P) Distributions of SON TSA-Seq percentiles for SPADs from DMSO-treated log phase (O) or G1 phase (P) cells compared with biological replicates, TPL-treated, or DRB-treated cells. Box plots show the median (black line), 25th–75th percentiles (box), whiskers (within 1.5× IQR), outliers (diamonds), and individual SPAD values (gray dots). Images in each panel are at same magnification, scale bars = 5 μm.

We then mapped genome organization relative to NS in asynchronous RPE1 cells by SON TSA-Seq 2.0.^20,30 Despite strong transcriptional inhibition, SON TSA-Seq profiles were largely unchanged in asynchronous cells after both DRB and TPL treatment, relative to untreated or DMSO-treated controls (Figure 1D). A similar conservation of SON TSA-Seq patterns after TPL treatment was observed in asynchronous HCT116 cells (Figure S1A–C).

We next asked whether transcription is required to re-establish NS association after mitosis. hTERT-RPE1 cells were arrested in mitosis with nocodazole, collected by shake-off, treated with TPL, and released into G1 while maintaining transcriptional inhibition (Figure 1E). Three hours after release, >95% of mitotic cells had entered G1 and reformed interphase nuclei with typical NS morphology, although G1 entry was slightly delayed in TPL-treated cells. RNA Pol II Ser5 phosphorylation remained strongly reduced (Figure 1F–G), comparable to the levels in log-phase cells. SON TSA-Seq 2.0 performed 3 hours after mitotic release again showed similar profiles in TPL-treated and control G1 cells, matching closely the TSA-Seq of asynchronous cells (Figure 1D).

Correlation coefficients comparing data sets from control versus transcription-inhibited cells were comparable to those between biological replicates (Figure 1H–M). Percentile normalization further showed that SPADs, defined as the top 5% of SON TSA-Seq scores,¹⁸ were largely conserved across control and transcription-inhibited cells, and between log phase and G1 cells (Figure 1N–P).

Together, these results show that both maintenance and post-mitotic establishment of genome association with NS occur largely independently of active transcription.

COL1A1 BAC transgenes autonomously target to NS after random integration

Despite our previous report of a partial transcription dependence to the NS association of HSPA1 transgenes,³¹ prior findings for the COL1A1 locus¹⁷ together with our TSA-Seq observations establish transcription-independent NS-genome association as the prevalent pattern. Therefore, we next decided to dissect the transcription-independent NS-targeting of individual loci using BAC transgenes.

We first examined three endogenous loci (GAPDH, COL1A1, COL1A2) by DNA immuno-FISH. As predicted by TSA-Seq, GAPDH and COL1A1 were closely associated with NS in the three human cell lines, whereas COL1A2 showed close association only in fibroblasts (Figure 2A–B, Figure S2A). In HCT116 cells, the close NS association of COL1A1 and GAPDH loci was insensitive to DRB treatment (Figure S2B–C), consistent with previous results for the endogenous COL1A1 locus,¹⁷ while the distance distribution did not change for the COL1A2 locus.

Figure 2. BAC transgenes containing SPAD DNA autonomously target to NS.

(A-B) Visualization and quantification of distance distribution of endogenous GAPDH, COL1A1, and COL1A2 genomic loci (3D DNA FISH, red) relative to NS (green) and DAPI (blue). in indicated human cell lines (n = 104-151).

(C-D) Transcription-independent NS-targeting of COL1A1 endogenous locus and BAC transgenes. (C) 3D DNA FISH (red) of integrated human COL1A1 or mouse Dhfr BACs in NIH 3T3 cells compared to NS (green) and DNA (blue). (D) Effects of transcriptional inhibition on NS proximity for endogenous COL1A1 locus in BJ-5ta-hTert cells or multicopy BAC arrays in NIH 3T3 cells visualized by 3D DNA FISH compared with DMSO controls (n = 103-106).

(E) SON (top 4 tracks), lamin B1, and MKI67IP TSA-Seq enrichment profiles showing proximity of six BAC candidates (red highlights) to NS, lamina and nucleoli.

(F) Schematic of BAC retrofitting strategy.

(G) HCT116 mixed cell populations stably transfected with indicated control vector or retrofitted BACs visualized by TetR-EGFP. NS (red) and DAPI (blue).

(H-I) Stacked histograms of showing fraction of foci per distance bin from NS (H) or nuclear lamina (I) (n=74-132).

(J-K) TetO foci distances to NS (J) and nuclear lamina (K) in HCT116 mixed populations represented as violin plots (n=74-132); significance tested by one-tailed z-test for fraction of foci within 0.25 μm, n.s., not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Images in each panel are at same magnification, scale bars, 2 μm (insets, 0.5 μm).

A multicopy COL1A1 BAC transgene array ³² also showed strong NS-association in NIH 3T3 cells (median distance 0.13 ± 0.01 μm), similar to the endogenous COL1A1 locus in human BJ fibroblasts (0.06 ± 0.01 μm) (Figure 2C–D). By contrast, a comparably sized multicopy DHFR BAC array,³² was positioned significantly farther from NS (0.33 ± 0.03 μm; p < 0.0001). NS-association of the COL1A1 BAC transgene array persisted after DRB and TPL treatment (Figure 2D), recapitulating the transcription-independent behavior of the endogenous COL1A1 locus (Figure S2B–C).

Thus, a COL1A1 BAC transgene containing ~165 kbp of the endogenous human COL1A1 locus autonomously targets to NS independent of transcription and provides a robust system to dissect NS-targeting mechanisms.

Autonomous Targeting to NS is a general property of BAC transgenes containing SPAD DNA

To determine whether autonomous NS-targeting is general property of SPADs, we examined six BAC transgenes containing ~100-200 kbp human DNA inserts. Single-copy BAC insertions were generated by random piggyBac (PB) transposition to provide a stringent test of autonomous targeting.

Four BACs derived from SPAD regions (GAPDH, ACTG1, RTEL1, and COL1A1 BACs) were compared with two BACs derived from Lamina-associated Domains (LADs) (OR51 and TPTE BACs) (Figure 2E). These SPAD BACs, named for representative genes contained within them, originate from gene-rich, early-replicating genomic loci positioned near maxima of broad SON TSA-Seq peaks, whereas the LAD BACs originate from a constitutive LAD (OR51) or from a LAD (TPTE) also with strong nucleolar association (MKI67IP TSA-Seq) (Figure 2E).

BAC vectors were retrofitted by recombineering with PB inverted terminal repeats (ITRs), a TetO-96mer array, and selectable markers (Figure 2F).³³ The retrofitting plasmid (Ctrl V1.0) was included as an internal control to establish the baseline targeting activity of the construct. After PB-mediated integration in HCT116 cells, mixed clonal populations containing independently integrated constructs were established (Figure 2F, Figure S2D). The integrated loci were visualized using TetR-EGFP binding to TetO arrays (Figure 2G).

All four SPAD BAC transgenes were positioned significantly closer to NS than LAD BACs and vector controls (Figure 2H&J) and comparably to the NS-associated endogenous COL1A1 and GAPDH loci in HCT116 cells (Figure 2B&J). Three of the four SPAD BAC transgenes showed reduced lamina proximity relative to the vector controls (Figure 2I&K). In contrast, the LAD BACs showed no significant change in NS proximity relative to the vector-only control but instead were positioned closer to the nuclear periphery (Figure 2G, I&K). These findings suggest repositioning of integrated SPAD BAC transgenes away from the nuclear periphery and towards the interiorly located _NS.6,34

In summary, these results demonstrate that autonomous NS-targeting is a common property of ~100-200 kbp fragments derived from SPAD regions.

Functionally redundant NS-targeting cis elements exist within the COL1A1 BAC

To identify cis elements responsible for NS-targeting, we dissected the COL1A1 BAC, which recapitulates the transcription-independent NS-association of the endogenous locus¹⁷ and has relatively low gene density among the examined SPAD BACs (Figure 2E&H). Because the endogenous COL1A1 locus showed its closest proximity to NS in fibroblasts of the three cell lines tested, we performed our assay in NIH 3T3 fibroblasts for maximum sensitivity. These cells also allow efficient BAC integration,³² and distinction of human transgene sequences from endogenous mouse loci.

A series of COL1A1 BAC fragments (11-159 kbp) spanning most of the 165 kbp genomic insert were generated (Figure 3A) through an iterative BAC recombineering approach³³ and stably integrated into NIH 3T3 chromosomes, again using PB transposition.

Figure 3. Cis elements mediate NS targeting in a functionally redundant manner, either with or without transcription.

(A) COL1A1 BAC schematic showing genes (red; arrowheads indicate direction of transcription), TetO-96mer array (green), and vector backbone (black). Solid lines mark fragments (sizes); Significant NS-targeting fragments (blue), not significant (red).

(B) Stacked histograms of TetO-NS distances for PB-mobilized empty vector or indicated BAC fragments in NIH 3T3 cells (n=97-105).

(C) Genome browser view of mouse chromosome 1 showing Lamin B1 and SON TSA-Seq profiles with regions selected (red highlights) for generating CRISPR-Cas9 docking lines.

(D) Schematic of site-specific integration of NS-targeting sequences in the COL1A1 BAC; only assay-relevant elements are shown.

(E) Schematic of tested fragments within the larger COL1A1 F17 fragment. Significant NS-targeting fragments (blue), not significant (red). Yellow highlight overlaps layered H3K27ac peaks across seven indicated cell types.

(F) Stacked histograms of TetO-NS distances for empty vector or BAC fragments integrated at the iLAD locus in NIH 3T3 cells (n=93-104).

(G) Genome browser view of four loci with H3K27ac profiles; highlighted peaks (yellow) were subcloned and tested. Previously tested fragments, F14 and F14 are also shown (brown highlights).

(H) Stacked histograms of TetO-NS distances after DMSO or DRB treatment for empty vector or BAC fragments integrated at the iLAD locus (n= 60-105). NS-targeting of DRB treated samples compared with corresponding DMSO controls.

Initial deletions from the 3’ end demonstrated that fragment F4 retained NS-targeting comparable to the near full-length F8 fragment (Figure 3A–B). The same 5’ deletions of fragments F4 and F8, which varied only in their 3’ ends, generated fragments F6 and F9 that retained similar NS-targeting activities. Additional truncations identified two non-overlapping ~11 kbp fragments, F14 and F17, each showing NS-targeting activity close to the near full-length BAC fragment (Figure 3B). In contrast, the 55 kbp F16 fragment covering most of the 3’ half of the COL1A1 BAC showed no NS-targeting activity above that of the vector-only control (Figure 3B).

Given the NS-targeting activities of the non-overlapping F14 and F17 fragments, both comparable in strength to the NS-targeting activity of the near full-length COL1A1 BAC, we conclude that the COL1A1 BAC contains at least two functionally redundant NS-targeting cis elements.

An ~600 bp region within the F17 COL1A1 fragment targets to NS

Since our aim was to identify NS-targeting cis elements, rather than completing the NS-targeting survey of the entire COL1A1 BAC, we next turned our attention to the further dissection of the F17 11 kbp fragment. While integration of 10s-100s of kbp of human DNA sequences within BAC transgenes can shield the influence of neighboring chromatin,^28,32,35,36 smaller DNA fragments can be influenced by chromosomal position effects. Therefore, we analyzed F17-derived fragments using site-specific integration at a defined genomic locus.

To achieve this, we used φC31 integrase-mediated recombination in NIH 3T3 fibroblasts (Figure 3C–D). Constructs containing attB sites were integrated into a pre-engineered docking site containing attP sequences and a TetO array located within a region between LADs (“iLAD” region) (Reg03) that lies at intermediate distances from both NS and the nuclear lamina (Figure S3A–D). Successful integration replaced selectable markers at the docking site, allowing enrichment of correctly targeted clones (Figure 3D). The full-length F17 fragment and smaller fragments derived from F17 (Figure 3E) were cloned into the integration vector and inserted into this docking site. Both PCR genotyping (Figure S3E) and FISH using a BAC probe flanking the targeted locus (Figure S3F) confirmed integration of the docking construct into the target locus. PCR genotyping also validated later attP/attB-mediated insertions into this docking site (Figure S3E). NS-targeting activity was then measured for each fragment.

Using this approach, the full-length F17 fragment showed NS-targeting comparable to that previously observed using random PB integration (Figure 3F). Analysis of progressively smaller fragments revealed several segments with significant NS-targeting activity that shared a common overlapping region (Figure 3E, yellow highlighted region, Figure 3F). F45, the smallest fragment (~600 bp) containing this region, showed NS-targeting activity only slightly lower but not statistically different from that of the parent F17 fragment. Larger fragments overlapping this region showed comparable targeting activity.

Thus, a ~600 bp sequence within the 11 kbp F17 is sufficient to reproduce most of its NS-targeting activity.

NS-targeting activity is a common property of several kbp cis elements with elevated H3K27ac

Within the 165 kbp COL1A1 human DNA BAC insert we identified two ~11 kbp fragments (F14 and F17) with NS-targeting activity and one 55 kbp fragment (F16) lacking such activity (Figure 3A–B). Both NS-targeting fragments overlap several kbp-wide peaks of H3K27ac present in multiple cell types, whereas the non-targeting F16 fragment lacks comparable enrichment (Figure 3G, brown highlighted region). The ~600 bp F45 fragment within F17 also coincides with a local H3K27ac peak (Figure 3E). A similar H3K27ac signal is observed at the syntenic mouse COL1A1 locus in CUT&Tag profiles of immortalized mouse embryonic fibroblasts (data not shown).³⁷

Because extended H3K27ac regions are characteristic of super-enhancers and enriched near SON TSA-Seq peaks,^18,38 we asked whether NS-targeting might be a common property of other H3K27ac-enriched cis elements. We therefore cloned three ~5-6 kbp fragments overlapping such H3K27ac peaks from BACs with NS-targeting activity: GAPDH F01 (5.8 kb), RTEL1 F01 (4.8 kb), and HSPA1 F01 (5 kb) (Figure 3G).

These fragments were integrated into the iLAD docking site in NIH 3T3 cells and tested for NS-targeting together with the COL1A1 F17 fragment. All four fragments showed significant NS-targeting relative to the vector control (Ctrl V2.0) (Figure 3H). The NS-targeting of the GAPDH and COL1A1 fragments was transcription-independent, consistent with the transcription-independent NS-targeting of their parent BACs (Figure S2B–C). The NS-targeting of the RTEL1 and HSPA1 F01 fragments instead were transcription-dependent, as had previously been described for the transcription-dependent targeting of HSPA1 plasmid and BAC transgenes.^31,39

We conclude that NS-targeting is a common property of several kbp cis elements enriched for H3K27ac modification and that such elements can mediate either transcription-dependent or transcription-independent NS association.

A subset of transcription factor motifs provides redundant NS-targeting activity to the F45 fragment

To identify cis regulatory elements responsible for the NS-targeting activity of the ~600 bp F45 fragment, we analyzed predicted transcription factor (TF) binding motifs using the JASPAR 2022 database (minimum score 500, p value < 10⁻⁵),⁴⁰ hosted at the UCSC genome browser.⁴¹ We identified five motif clusters corresponding to potential binding sites for eight TFs, including TFAP2A/B/C, MXI1, ZBTB18, TWIST1, ZNF384, and ZNF460 (Figure 4A). The first 5′ cluster contains overlapping motifs for TFAP2A, TFAP2B, TFAP2C, and MXI1. A second cluster in the middle of the fragment contains overlapping motifs for ZBTB18 and TWIST1, followed by a single ZNF384 binding site. Toward the 3′ end, a ZNF460 site is present, followed by a cluster of overlapping ZNF384 motifs.

Figure 4. Transcription factor binding motifs within the ~600 bp COL1A1 BAC F45 fragment confer NS-targeting activity.

(A) Schematic of the 600 bp F45 fragment with predicted transcription factor (TF) binding sites. Motifs (brown stars) with orientation arrows are shown. Mutant F45 constructs used to dissect NS-targeting are indicated. Blue TFs conferred NS-targeting; red TFs did not.

(B) Stacked histograms of TetO-NS distances for empty vector or indicated constructs integrated at the iLAD locus in NIH 3T3 cells.

(C-E) ChIP-qPCR enrichment of H3K27ac and endogenous ZNF384 at F45 fragments integrated at iLAD locus. Yellow highlights in (C) mark reference target loci, based on available MEF H3K27ac ChIP-Seq data. Error bars show mean ± SEM from two biological replicates normalized to IgG.

(F) Schematic of the tethering assay. TetR-EGFP-TF fusion proteins are recruited via TetO-TetR interaction.

(G) Maximum-intensity projections of three optical z-sections from representative nuclei expressing TetR-EGFP-TF with TetO foci (white arrowheads) visualized by TetO FISH (red) at the iLAD locus. Insets demonstrate proximity of TetO foci to NS (green). Images in each panel are at same magnification, scale bars= 5 μm (insets 0.5 μm).

(H) Stacked histograms of TetO-NS distances for indicated TetR-EGFP-TF fusion proteins recruited to the iLAD locus in NIH 3T3 cells (n= 105-115).

To test whether these motifs mediate NS-targeting, we generated a mutated (“all-mut”) F45 fragment in which all predicted TF binding motifs were disrupted (Figure 4A). To minimize potential effects on nucleosome positioning or spacing of nearby regulatory elements, nucleotide substitutions rather than deletions were introduced within the core TF binding motifs (Figure S4). After site-specific integration into the iLAD docking site, this “all-mut” fragment showed NS association indistinguishable from the vector control (Ctrl V2.0) (Figure 4A–B), indicating that one or more of the predicted TF binding sites are required for NS-targeting.

We next restored each motif cluster individually within the all-mut background (restored Site Directed mutate Motifs (rSDMs): rSDM1-rSDM5), generating fragments containing a single intact TF binding site or cluster (Figure 4A). Restoration of the 5′ motif cluster containing TFAP2A/B/C and MXI1 failed to rescue NS-targeting. In contrast, restoration of each of the remaining four motif clusters restored NS-targeting activity to levels comparable to the wild-type F45 fragment (Figure 4B). These functional motifs correspond to predicted binding sites for ZBTB18, TWIST1, ZNF384 and ZNF460. Notably, the rSDM3 fragment containing a single ZNF384 binding motif showed NS-targeting comparable to rSDM5, which contains 11 overlapping ZNF384 motifs.

In summary, four distinct TF binding motif sites within the F45 fragment are individually sufficient to confer NS-targeting activity, demonstrating strong functional redundancy among these cis elements.

A subset of the TFs predicted to bind the F45 fragment show NS-targeting activity in a tethering assay

In the preceding section we identified TF binding motifs that individually restored the NS-targeting activity of the ~600 bp F45 fragment. In some cases, multiple TFs were predicted to bind the same or overlapping motifs, which are similar for both mouse and human TF orthologs.⁴⁰ The mouse orthologs of all TFs predicted to bind are expressed in NIH 3T3 cells, with TFAP2A/B/C expressed at low levels (34-36 percentile) and MXI1, TWIST1, ZBTB18, and ZNF384 more highly expressed (92-99 percentile).⁴² We observed enrichment of H3K27ac marks on this WT F45 fragment, but not on “all-mut” F45 fragment, indicating that the recruitment of TFs on the identified sites lead to establishment of this histone mark (Figure 4C–D). Additionally, ChIP-qPCR validated recruitment of endogenous mouse ZNF384 TF to the cognate motifs located on the WT F45 fragment, but not on “all-mut” F45 fragment integrated within the iLAD docking cell line (Figure 4E).

To determine whether NS-targeting activity could be attributed to individual TFs, we applied a TetO/TetR tethering assay to test each TF individually (Figure 4F). All TetR-EGFP-TF fusion proteins were expressed at similar levels. Our attP/attB site-specific recombination approach introduced a total of 192 TetO repeats at the docking locus (96 introduced during the docking site creation and another 96 after attB insertion; Figure 3D). The insertion of the second 96mer TetO repeat facilitated the visual validation of cell clones with true site-specific integration of the attB plasmid. TetR-EGFP served as a negative control whereas TetR-EGFP fused to full-length p53 served as a positive control, based on previous reports of p53-mediated NS-targeting.²⁴

One technical difficulty was that many of the TetR-EGFP-TF fusion proteins showed non-diffuse, localized nucleoplasmic distributions, even in the absence of any TetO repeat, obscuring identification of the docking site. To circumvent this issue, we used DNA FISH to visualize the TetO repeats when testing these TetR-EGFP-TF fusion proteins (Figure 4G).

Using the iLAD docking site (Figure S3A), tethering of TetR-EGFP-p53 produced a significant increase in NS proximity relative to the TetR-EGFP control (Figure 4F–H), validating the assay. Tethering of full-length TWIST1, ZBTB18, ZNF384, and ZNF460 similarly resulted in significant NS-targeting (Figure 4H). In contrast, tethering of TFAP2A, TFAP2B, TFAP2C, or MXI1 showed no statistically significant difference from the TetR-EGFP negative control.

Thus, using this TetO/TetR tethering assay, each TF predicted to bind the four functional motif sites within the F45 fragment conferred NS-targeting activity comparable to p53, whereas the four TFs predicted to bind the non-targeting rSDM1 motif cluster did not.

NS-targeting is a common property of acidic activation domains including those within the acidic TFs predicted to confer NS-targeting to the F45 fragment

Previous studies suggested that transcription factors may target to NS through a common proline-rich domain (PRD) based on observations that mutations in the p53 PRD reduce NS-targeting.^24,43 However, we observed no correlation between the presence of PRDs and the NS-targeting activities among the eight TFs predicted to bind the F45 fragment. Instead, the four NS-targeting TFs (TWIST1, ZBTB18, ZNF384, ZNF460) predicted to bind to the DNA motifs that confer NS-targeting activity to the F45 fragment all contain acidic activation domains (AADs), whereas none of the four non-targeting TFs contain AADs.

AADs are intrinsically disordered activation domains enriched in acidic residues flanking clusters of hydrophobic amino acids, particularly aromatic and leucine residues.^44–47 Only ~8% of human TFs have been formally annotated as containing AADs, although the actual percentage of TFs containing AADs is predicted to be higher, as the hydrophobic amino acids within AADs do not have a strict sequence grammar or motif-based distribution.^48,49 Our laboratory previously demonstrated that tethering the strong VP16 AAD induced large-scale chromatin decondensation of a heterochromatic, gene-amplified chromosome region,⁴⁶ and long-range, directional movement of a plasmid transgene from the nuclear periphery towards the nuclear interior.⁵⁰ These activities were shared by several AADs, including a synthetic peptide ‘DELQPASIDP’ (“DELQP” peptide), but not by glutamine-rich or proline-rich non-acidic transcriptional activation domains (NAADs).^51,52 Recruitment of the VP16 AAD, p65 AAD, DELQP peptide, and VP64, a tetramer of the VP16 AAD, can also reposition endogenous genomic loci from the nuclear periphery to the interior.^53,54 Notably, recruitment of VP16 also induced clustering of NS around the heterochromatic, gene-amplified chromosome region.⁴⁶

We therefore tested whether NS-targeting is a common property of AADs. Using the TetO/TetR tethering system, we examined several previously characterized AADs (VP16, p65 AAD1, p65 AAD2, DELQP peptide, and p53) alongside glutamine-rich (SP1, OCT1, OCT2) and proline-rich (AP-2A/TFAP2A) NAADs, as well as the N-terminal domain of CTCF (Figure 5A, Figure S5), which had been shown to contain similar large-scale chromatin decondensation activity as VP16.^50–53,55 TetR-EGFP AAD and NAAD fusion proteins were transiently expressed in NIH 3T3 cells containing the Reg03 iLAD docking site tagged with a TetO 96-mer array.

Figure 5. Acidic Activator Domains (AADs) show chromatin context-dependent NS-targeting activity.

(A) Representative plots of predicted acidic transcription activation domains (AADs) or non-acidic activation domains (NAADs) in indicated transcription factors. Regions tested in TetO-TetR recruitment assay are shaded (gray); red lines mark the AAD prediction threshold.

(B-C) Stacked histograms of TetO-NS distances of TetR-EGFP fusion proteins recruited to the iLAD locus (B) or LAD docking sites (C) in NIH 3T3 cells (n= 61-123).

Strikingly, both the VP16 AAD and the DELQP peptide produced strong NS-targeting relative to the TetR-EGFP control (VP16: 71% <0.25 μm from NS; DELQP: 72% <0.25 μm from NS). Other AADs (p65 AAD1, p65 AAD2, p53 AAD) also produced statistically significant NS-targeting (Figure 5B). In contrast, none of the tested NAADs or the N-terminal domain of CTCF showed significant NS-targeting.

Having established that AADs commonly confer NS-targeting, we next asked whether the AADs within the acidic TFs identified in our F45 fragment analysis also share this activity. AAD-like sequences within ZBTB18, ZNF384, ZNF460, and TWIST1 were identified using the ADpred deep-learning model for activation domain prediction,⁴⁷ based on high ADpred scores (≥0.8 value over ≥10-15 residues) (Figure 5A, Figure S5A, grey highlights).

Consistent with these predictions, all (8/8) tested AAD-like peptides with AAD-prediction scores well above the previously established AAD cut-off score⁴⁷ (VP16, p65 AAD1 & AAD2, p53, ZBTB18, ZNF384, and ZNF460), as well as close to the AAD-prediction cutoff score (TWIST1), showed robust NS-targeting activity in the tethering assay (Figure 5B). In contrast, all (5/5) peptide sequences with low predicted AAD scores, including the TFAP2A peptide, showed no NS-targeting activity. This was consistent with the NS-targeting observed for full-length TFs (Figure 4E).

Together, these results demonstrate that NS-targeting is a common property of AADs. Moreover, among TFs predicted to bind the NS-targeting F45 fragment, only those containing AAD-like sequences exhibited NS-targeting activity, with AAD peptides producing targeting comparable to their full-length parent TFs.

NS-targeting activity of AADs is chromatin context-dependent

Whereas in the preceding Section we showed how recruitment of AADs at the Reg03 iLAD docking site induces NS-targeting, previously we had not noticed an obvious NS-association of plasmid transgenes after their repositioning from the nuclear periphery to interior.⁵⁰ These apparently conflicting results prompted us to investigate whether the NS-targeting of AADs might depend on the chromatin context by comparing the NS-targeting activities of AADs and NAADs tethered to TetO arrays at the Reg03 iLAD docking site versus at the Reg01 docking site integrated into a large LAD (Figure S3A).

NS-targeting was significantly reduced for most AADs when tethered within the LAD compared to the iLAD (Figures 5B–C). Although NS-targeting remained statistically significant for several AADs (VP16, DELQP peptide, p65 AAD2, TWIST1, ZBTB18, ZNF384, ZNF460 AAD1 and AAD2), its magnitude was consistently diminished, while the p53 AAD and p65 AAD1 showed no statistically significant NS-targeting relative to the TetR-EGFP control.

In conclusion, these results indicate that although NS-targeting is a general property of AADs, its magnitude depends on chromatin context. This observation predicts that there will not be a direct correlation between the binding of acidic transcription factors across the genome and NS proximity.

NS-targeting by AADs is transcription-dependent

Because we previously identified both transcription-dependent and transcription-independent NS-targeting elements (Figure 3G–H), we next tested whether the NS-targeting activity of tethered AADs requires active transcription. We compared the NS-association after TetO/TetR-mediated tethering of the two strongest AAD constructs, VP16 and the DELQP peptide, at both the Reg03 iLAD and Reg01 LAD docking sites with and without the TPL transcription inhibitor (Figure 6A–B).

Figure 6. Some acidic transcription factors show transcription-dependent NS-targeting via acidic activator domains (AADs) and transcription-independent NS-targeting through additional domains.

(A-C) Stacked histograms of TetO-NS distances after DMSO or TPL treatment for TetR-EGFP AAD fusions recruited to the NIH 3T3 iLAD (A) or LAD locus (B) (n= 54-123), or for TetR-EGFP-p53 (full-length) recruited to the iLAD locus (C) (n= 57-106).

(D) Schematic of TF domains and deletion constructs. Dotted line indicates deleted residues; double-headed arrows mark overlapping domains in ZNF460. Numbers denote amino acid boundaries.

(E) Stacked histograms of TetO-NS distances after DMSO or DRB treatment for TetR-EGFP fusion proteins recruited at the NIH 3T3 iLAD locus (n= 96-105).

(F) Summary schematic of NS-targeting by TetR-EGFP constructs at the iLAD locus after DMSO, DRB, or TPL treatment. Solid colors indicate significant NS-association; lighter shades indicate loss of NS-targeting.

NS association after TPL was compared with DMSO controls and with TetR-EGFP negative controls treated under the same conditions. TetR-EGFP alone showed a small but statistically insignificant reduction in NS proximity after TPL treatment at both loci (Figure 6A–B), possibly reflecting weak transcription-dependent NS-targeting from the selectable marker present at the docking sites. In contrast, TPL eliminated the NS-targeting activity relative to the TetR-EGFP control of both the VP16 AAD and DELQP peptide TetR fusion proteins (Figure 6A–B).

Thus, NS-targeting conferred by two of the strongest AADs is transcription-dependent in both iLAD and LAD contexts.

Some acidic transcription factors contain additional domain(s) that confer transcription-independent NS-targeting activity

The transcription-dependent NS-targeting of AADs creates an apparent paradox. Our TSA-Seq experiment showed that most genomic regions establish and maintain their NS-association even after transcriptional inhibition (Figure 1); moreover, COL1A1 BAC transgenes recapitulated this transcription-independent NS-targeting (Figures 2 and 3). Dissection of the COL1A1 F45 fragment established NS-targeting activity conferred by the DNA motifs predicted to bind acidic TFs, and tethering of the AADs from these acidic TFs led to NS targeting (Figures 4 and 5). However, the NS-targeting activity of two strong AADs was transcription-dependent, presenting the question of whether the NS-targeting activities of the AADs of the acidic TFs predicted to bind the COL1A1 F45 fragment were also transcription-dependent and, if so, whether additional domains within acidic TFs confer transcription-independent NS-targeting.

While this work was in progress, a separate study suggested that a proline-rich domain (PRD) in p53 contributes to NS-targeting.²⁴ To test whether such a domain might confer transcription-independent NS-targeting by p53, we examined the NS-targeting activity of full-length p53 using TetO/TetR tethering combined with DNA FISH to visualize the TetO locus. After TPL treatment, the NS-targeting activity of the TetR-EGFP fused to full-length p53 was significantly reduced relative to the DMSO control and became indistinguishable from the TetR-EGFP negative control (Figure 6C). Thus, NS-targeting by p53 is also transcription-dependent.

We next examined ZNF384 and ZNF460, which showed the strongest NS-targeting activity among the TFs predicted to bind the F45 fragment. To determine whether transcription-independent NS-targeting might arise from domains outside their AADs, we compared the NS-targeting activities of fusion proteins containing full-length (FL) ZNF384 or ZNF460 TFs with the NS-targeting activities of truncated fusion proteins with the AADs of these TFs deleted or with only the TF AADs included (Figure 6D–E). Because ZNF460 has two AADs, we used a larger deletion that eliminates both AADs (Figure 6D). Both FL ZNF384 and the truncated ZNF384 with deletion of its AAD (ΔAAD ZNF384) showed NS-targeting activity that was unchanged after transcriptional inhibition by DRB. In contrast, the NS-targeting activity of the ZNF384 AAD alone was reduced after transcriptional inhibition by DRB (Figure 6E–F). Similar results were observed for ZNF460, for which deletion of both AADs produced NS-targeting that was insensitive to DRB, whereas each AAD showed reduced NS-targeting after transcriptional inhibition by DRB (Figure 6E–F).

In conclusion, both ZNF384 and ZNF460 contain additional domains that show a transcription-independent NS-targeting activity, distinct from the transcription-dependent activity of their AADs.

DISCUSSION

The proximity of highly active genes to nuclear speckles (NS) led to a transcription-driven model of NS association,⁵⁶ in which transcription-dependent accumulation of splicing factors and RNA-binding proteins (RBPs) over transcripts produced by active genes nucleates NS formation.^3,25,57,58 However, genome-wide analyses using SPRITE, TSA-Seq, multiplexed FISH, and super-resolution imaging show that only a subset of highly active genes are stably NS-associated,^18,19,21,22 while many active loci remain distal to NS.³⁴ Conversely, the COL1A1 locus remains NS-associated even after transcriptional inhibition,¹⁷ indicating that factors beyond transcription contribute to genome positioning relative to NS.

Therefore, to identify determinants of the reproducible positioning of Speckle Associated Domains (SPADs), we applied a top-down unbiased strategy. Genome-wide TSA-Seq analyses revealed that active transcription is not required either to maintain NS-chromatin associations during interphase or to reestablish associations in early G1 nuclei after mitosis. In contrast to previous models,^3,25,56 these findings indicate that the close NS-association of gene-dense SPADs that are 100s to 1000s of kbp in length is not simply a consequence of their high transcriptional activity. This raises a fundamental question: if transcription is dispensable, what mechanisms determine and/or redundantly reproduce basal NS-targeting?

To address this question, we first demonstrated the NS-targeting activities of ~150-200 kb sequences derived from the centers of SPADs after their random chromosomal integration. We then followed this with a systematic dissection of the NS-targeting activities of DNA sequences from a single ~165 kb human DNA sequence containing the COL1A1 gene that aligned with a prominent TSA-Seq subpeak within a multi-Mbp SPAD.

Our results show that NS-targeting is not driven by large contiguous genomic regions but instead by multiple smaller cis regulatory elements (CREs) that function redundantly within SPADs. Two smaller, ~11 kb NS-targeting fragments identified within the COL1A1 BAC corresponded to genomic sites enriched in H3K27ac. Integration of several 5-6 kb DNA fragments overlapping such H3K27ac peaks from other genomic loci established that NS-targeting is a common property of sequences containing broad H3K27ac peaks characteristic of super-enhancers.^18,38

This redundancy of NS-targeting extends to even smaller scales. An ~600 bp COL1A1 BAC fragment contained multiple acidic TF binding motifs, each individually sufficient to induce NS-targeting comparable to the wild-type 600 bp fragment. All four of the acidic TFs predicted to bind to these NS-targeting motifs possessed AADs with NS-targeting activity. Moreover, two of these four acidic TFs exhibited two separable NS-targeting activities located in different protein regions: a transcription-dependent NS-targeting activity localized to their AADs and a transcription-independent NS-targeting activity localized elsewhere in a subset of acidic TFs. More generally, NS-targeting activity appears to be a general property of AADs, as all AADs tested- including VP16, p65 and p53 AADs- showed NS-targeting activity that was not seen for several NAADs.

Such redundancy of NS-targeting also explains the fine-tuning of NS-association by additional TF binding events. For example, activation of p53 and HIF2α induces their binding to a limited set of genomic sites, adding their NS-targeting activities to loci already occupied by constitutively bound TFs. ^24,43 This results in modest repositioning of target genes closer to NS and a concomitant increase in transcription. Even small positional shifts relative to NS can yield multi-fold increases in gene expression through amplification associated with nuclear speckle contact.^20,24,43,59 Notably, we found that the NS-targeting activity of tethered acidic TFs is strongly chromatin-context dependent, with reduced NS-targeting within LAD compared to iLAD regions. This may help explain why genes moving closer to NS after HSF1, p53, or HIF2α induction are frequently located in genomic regions already prepositioned near NS.^20,24,43

The complete spectrum of protein domains capable of mediating NS-targeting remains unidentified. A previous study attributed NS-targeting activity of p53 exclusively to its proline-rich domain, as its deletion reduced targeting while AAD mutations disrupting transcriptional activation did not.²⁴ That study did not test the p53 AAD directly, presumably because its NS-targeting activity was assumed to depend on its transcriptional activation activity. In contrast, our results show that the NS-targeting activity of an AAD can be uncoupled from its transcriptional activation activity. For example, the AAD-like DELQP peptide strongly targets to NS yet lacks detectable transcriptional activation activity. Other proteins containing proline-rich domains have also been proposed to mediate NS-targeting, including HIF2α.⁴³ Of the eight TFs predicted to bind the COL1A1 600 bp fragment, all four acidic TFs exhibited NS-targeting activity, although only one (ZNF384) contained a predicted proline-rich speckle targeting motif.⁴³ Conversely, among four non-acidic TFs lacking NS-targeting activity in our assays, three contained predicted proline-rich motifs (TFAP2A/B/C).⁴³

Additional proteins implicated in NS association include MAZ, CTCF, and cohesin, potentially through a proline-rich speckle targeting motif within RAD21.^43,60 Defining NS hubs through inter-chromosomal Hi-C interactions revealed that NS-association correlates more strongly with trans contacts at constitutive super-enhancers than with H3K27ac levels.⁶⁰ This observation is consistent with our findings that 4/4 tested super-enhancer like DNA fragments containing broad H3K27ac peaks possess NS-targeting activity, although not all H3K27ac-enriched regions confer such targeting. Prior analyses also identified enrichment of MAZ at NS-proximal regions, whose depletion decreased NS interactions at strongly NS-associated loci while increasing interactions at weaker loci.⁶⁰

The contribution of CTCF and cohesin to NS positioning relative to acidic TFs remains unclear.⁴³ Microscopy-based studies indicate that CTCF- or cohesin-mediated changes in NS proximity are generally small and detectable primarily for loci already located near NS. In contrast, AADs can induce long-range repositioning from NS-distal regions near the nuclear periphery toward NS-proximal locations in the nuclear interior.^50,53 Here we showed NS-targeting activity of stronger AADs even when they were tethered to LADs positioned at the nuclear periphery. Furthermore, many proposed NS-targeting effects of CTCF or cohesin were inferred indirectly through SON CUT&RUN or CUT&Tag assays that directly measure local enrichment of speckle components over DNA rather than distance to NS.^27,43 In our assays, we did not observe NS-targeting activity for the N-terminal domain of CTCF, despite its ability to promote chromatin decondensation similar to the VP16 AAD⁵¹ and to recruit cohesin.⁶¹

In summary, our findings establish a multifactorial functionally redundant framework in which intrinsic CREs and associated proteins cooperatively position specific genomic loci around NS. This “Velcro-like” multifactorial NS-targeting model (Figure 7) explains both the robustness of NS-genome interactions across cell types and their plasticity in facilitating rapid gene regulation.

Figure 7. Cooperative and redundant mechanisms anchoring genomic loci at the NS periphery.

(A) Schematic illustrating how partially redundant factors position genomic loci at the NS periphery. (B) AADs drive transcription-dependent targeting, while additional TF domains and other unidentified elements mediate transcription-independent recruitment. This cooperative recruitment contributes to stable anchoring at NS periphery.

Limitations of the Study

Several key questions remain. Although NS-targeting correlates strongly with H3K27ac enrichment, only a subset of hyperacetylated regions showed targeting activity. Future experiments will be required to test whether reduction of H3K27ac alters NS association. More broadly, general mechanisms of transcription-dependent and transcription-independent NS-targeting remain unclear. Recruitment of AADs can induce long-range chromosome movements,⁵⁰ raising the possibility that these movements facilitate NS targeting. Indeed, as shown in a companion study, transcriptional inhibition disrupted stable NS-anchoring of the HSPA1 gene and transgenes while inducing recurrent long-range movements to and from NS dependent on the acidic HSF1 TF.³⁹ If true, this would imply that a distinct mechanism provides stable NS anchoring or “sticking”, either through nascent RNAs at flanking expressed genes and/or through cooperative recruitment of additional proteins containing “sticking” domains, possibly including proline-rich NS-targeting motifs.^24,43 NS-anchoring conferred by nascent transcripts attached to flanking chromosomal regions would explain the transcription-dependent NS-targeting activity of AADs, including the transcriptionally inactive DELQP AAD-like peptide. Such mechanisms could explain the strong chromatin-context dependence of NS-targeting described here and in previous studies examining the beta-globin locus.⁵⁷ Finally, we speculate that additional NS-targeting proteins or domains may also exist beyond the acidic TFs identified here.

Understanding the functional consequences of NS proximity for gene regulation in health and disease is another important direction. NS-targeting by p53 and HIF2α has been studied extensively.^24,43 Both ZNF384 and ZNF460, identified in our COL1A1 BAC dissection as NS-targeting TFs, have been associated with cancer. ZNF384 scaffolds DNA damage repair complexes,⁶² and regulates genes, including COL1A1,⁶³ matrix metalloproteinases (MMPs)⁶⁴ and cyclin D1.⁶⁵ ZNF384 is overexpressed in acute leukemias and enriched at promoters, enhancers and TAD boundaries.⁶⁶ Moreover, the presence of over nineteen oncogenic ZNF384 fusion proteins involving partners such as EP300, TAF15, CREBBP, or ARID1B define a subtype of B-cell acute lymphoblastic leukemia (B-ALL).⁶⁷ These fusions may alter NS-targeting by misdirecting epigenetic regulators or reposition to NS genomic loci bound by ZNF384 fusion partners, thereby promoting aberrant transcriptional programs. ZNF460 is broadly expressed at low levels in human cells.⁶⁸ Elevated ZNF460 expression has been associated with acute myeloid leukemia, gastric cancer, and colon cancer, and has been linked to the epithelial-to-mesenchymal transition.^68,69

Finally, our recent work identifying perispeckle networks as additional niches for active transcription⁷⁰ suggests a possible continuum of NS targeting. We speculate that weaker NS-targeting DNA elements may position loci at low frequency to NS but at higher frequency to these perispeckle networks away from NS. Indeed, a lower number of super-enhancers map near weaker, “Type II” SON TSA-Seq peaks, compared to the greater number of super-enhancers observed near SPADs.¹⁸ Future studies will be needed to dissect these gradations of targeting, including contributions from expressed RNAs and cooperative factors. Such work should help distinguish stable NS anchoring from dynamic repositioning within the broader perispeckle environment and clarify how these mechanisms contribute to gene regulation.

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to the corresponding author, Andrew Belmont (asbel@illinois.edu).

Materials Availability

Details on antibodies/reagents used in the study have been included in the Key Resources Table. The reagents and constructs are available from the corresponding author upon reasonable request. Sequence details for all the constructs used in the study have been submitted to GenBank and corresponding accession numbers of constructs are provided in Table S1.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-SON	Custom made against hSON peptide (947–964; Cys-DPYRLGHDPYRLTPDPYR)	Pacific Immunology Corp.
Anti-Lamin B1 antibody	Abcam	ab16048
Anti-MKI67IP antibody	Atlas Antibodies	HPA035735
GFP Monoclonal Antibody (3E6)	ThermoFisher Scientific	A-11120
Anti-RNA Pol II Ser5P monoclonal antibody, clone CTD4H8	Millipore-Sigma	05-623
Acetyl-Histone H3 (Lys27) (D5E4) Rabbit Monoclonal Antibody	Cell Signaling Technology	8173S
ZNF384 Polyclonal Antibody	ThermoFisher Scientific	PA5-52044
Normal Goat IgG (for ChIP assay)	Cell Signaling Technology	2729S
Normal goat serum	Millipore-Sigma	G9023
Alexa Fluor® 488 AffiniPure Goat Anti-Mouse IgG	Jackson Immunoresearch Inc.	115-545-146
Alexa Fluor® 647 AffiniPure Goat Anti-Mouse IgG	Jackson Immunoresearch Inc.	115-605-174
Alexa Fluor® 594 AffiniPure Goat Anti-Rabbit IgG	Jackson Immunoresearch Inc.	111-585-003
Fluorescein (FITC) AffiniPure® Goat Anti-Rabbit IgG (H+L)	Jackson Immunoresearch Inc.	111-095-144
Rhodamine Red-X (RRX) AffiniPure™ Goat Anti-Rabbit IgG	Jackson Immunoresearch Inc.	111-296-003
Alexa Fluor® 647 AffiniPure® Goat Anti-Rabbit IgG (H+L)	Jackson Immunoresearch Inc.	111-605-144
Streptavidin, Alexa Fluor™ 594 Conjugate	ThermoFisher Scientific	S11227
Alexa Fluor® 647 Streptavidin	Jackson Immunoresearch Inc.	016-600-084
Alexa Fluor® 594 AffiniPure® Goat Anti-Mouse IgG (H+L)	Jackson Immunoresearch Inc.	115-585-146
Alexa Fluor® 647 IgG Fraction Monoclonal Mouse Anti-Digoxin	Jackson Immunoresearch Inc.	200-602-156
Chemicals, peptides, and recombinant proteins
Janelia Fluor® 646 HaloTag® Ligand	Promega	GA1120
Triptolide	Millipore-Sigma	T3652
DRB	Millipore-Sigma	D1916
Nocodazole	Millipore-Sigma	M1404
DMSO	Millipore-Sigma	D2650-5x10mL
Fibronectin	Millipore-Sigma	F1141
Bovine Serum Albumin	Millipore-Sigma	A7906
5-Ethynyl uridine (EU)	Jena Bioscience	CLK-N002-10
Digoxigenin-11-dUTP, alkali-stable	Millipore-Sigma	11093088910
Biotin-14-ATP	ThermoFisher Scientific	19524016
Glass coverslips #1.5, 12mm diameter	Fisher Scientific	12-545-81
Mouse Cot-1 DNA	ThermoFisher Scientific	18440016
Human Cot-1 DNA	ThermoFisher Scientific	15279011
G-418 Sulfate	Fisher Bioreagents	BP673-1
Zeocin Selection Reagent	ThermoFisher Scientific	R25001
Chloramphenicol	Millipore-Sigma	220551-25GM
Blasticidin S HCl	ThermoFisher Scientific	R21001
Hygromycin B (50 mg/mL)	ThermoFisher Scientific	10687010
Puromycin	Gibco	A11138-03
Polybrene	Millipore-Sigma	TR-1003-G
Lipofectamine 2000	ThermoFisher Scientific	11668019
Formamide, Deionized	Millipore-Sigma	S4117
5-Bromo-2′-deoxyuridine (BrdU)	Millipore-Sigma	B5002-250MG
5-Bromo-2′-deoxycytidine (BrdC)	ThermoFisher Scientific	AAJ6545603
bisBenzimide H 33258 (Hoechst 33258)	Millipore-Sigma	B2883-25MG
Exonuclease III (E.Coli)	New England Biolabs	M0206S
Roche cOmplete Mini EDTA-free tablets	Millipore-Sigma	11836170001
ChIP-Grade Protein G Magnetic Beads	Cell Signaling Technology	9006S
RIPA Buffer (10X)	Cell Signaling Technology	9806S
ChIP Wash Buffer	Santa Cruz Biotechnology	sc-45002
Tamoxifen	Millipore-Sigma	T5648
DAPI	Millipore-Sigma	D9542
Paraformaldehyde	Millipore-Sigma	P6148-500G
Bacterial strains
DH5a Escherichia coli strain	New England Biolabs	C2984I
SW102	Published earlier.33	E. Coli recombineering strain
Experimental models: Cell lines
HCT116 (human colon carcinoma)	4D Nucleome Consortium	ATCC-CCL-247
hTERT RPE1 (human retinal pigment epithelial)	4D Nucleome Consortium	ATCC CRL4000
NIH 3T3 (mouse fibroblast)	ATCC	CRL-1658
HEK293T cells	ATCC	CRL-3216
Oligonucleotides
Oligonucleotides	Integrated DNA Technologies (IDT)	See Table S1
Recombinant DNA
pMini-HB-attB-Neo-Puro-HSVtk-PB (hereafter Ctrl V1.0 vector)	See Table S1	This study
pMini-HB-attB-Neo (hereafter Ctrl V2.0 vector)	See Table S1	This study
pX330A-1x3	Addgene	Plasmid # 58767
FRT-PGK-FRT-attP-Puro-DTA plasmid	Gifted by David Gilbert, SDBRI	N/A
attB-Neo-SacB	Gifted by David Gilbert, SDBRI	N/A
pJoyC030	Gifted by Bas van Steensel, NKI	N/A
mPB-L3-ERT2-tatRRR-mCherry	Gifted by Bas van Steensel, NKI	N/A
CTD-318A47 (human RTEL1 BAC)	ThermoFisher Scientific	Catalog #: 96012, Clone Id: 3184A7
CTD-2269C15 (human COL1A2 BAC)	ThermoFisher Scientific	Catalog #: 96012, Clone Id: 2269C15
RP11-765O14 (human ACTG1 BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 765O14
RP11-369N23 (human GAPDH BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 369N23
RP11-267M22 (human COL1A1 BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 267M22
RP11-92G8 (human HSPA1 BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 92G8
RP11-351B23 (human TPTE BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 351B23
RP11-715G8 (human OR51 BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 715G8
RP11-1136P2 (human RNPS1 BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 1136P2
RP11-634L10 (human ARHGDIA BAC)	ThermoFisher Scientific	Catalog #: RPCI11.C, Clone Id: 634L10
RP23-99P16 (mouse BAC)	BACPAC Genomics Inc	Clone Id: RP23-99P16
CITB-057L22 (mouse Dhfr/Msh3 BAC)	Published earlier.32	Gifted by Edith Heard (Curie Institute, Paris, France)
Critical Commercial Assays
End-It DNA End-Repair Kit	Epicentre Biotech	ER0720
TruSeq ChIP Library Preparation Kit – Set A	Illumina	IP-202-1012
NEB® PCR Cloning Kit	New England Biolabs	E1203S
QIAGEN Large Construct Kit	QIAGEN	12462
Dneasy Blood & Tissue kit	QIAGEN	69504
QIAGEN Plasmid Midi Kit	QIAGEN	12143
QIAprep Spin Miniprep Kit	QIAGEN	27106
Deposited data
RPE1 SON TSA-Seq (G1 phase cells)	This study	Accession No. 4DNESKYNUNSG, 4DNESRI54D3Q, 4DNESGE6ZI4N, and 4DNESJ4OXHI8
RPE1 SON TSA-Seq (log phase cells)	This study	Accession No. 4DNESNXH2OHP, 4DNESF79UZDD, 4DNESANW8A1W, and 4DNES3LO4836
HCT116 SON TSA-Seq (used in Figure S1)	This study	Accession No. 4DNESQQ92325 and 4DNESUTJFN2R
NIH 3T3 SON TSA-Seq	This study	Accession No. 4DNESPRABC27
NIH 3T3 LB1 TSA-Seq	This study	Accession No. 4DNES8Y37MM4
HCT116 SON TSA-Seq	Published earlier.20	Accession no. 4DNFI9GKAT7E
HFF SON TSA-Seq	Published earlier.20	Accession no. 4DNFIW688BVB
HCT116 LB1 TSA-Seq	Published earlier.73	Accession no. 4DNESTE8NJOE
HCT116 MKI67IP TSA-Seq	Published earlier.73	Accession no. 4DNESAHA7E69
H3K27ac ChIP-Seq on mouse embryonic fibroblast (MEF)	Published earlier.74	Accession no. ENCSR000CDI
NIH 3T3 RNA-Seq	Published earlier.42	Accession no. GSM5388182
Image source data for Figure images	This study	doi: 10.17632/rjwkxm566p.1
Softwares and algorithms
SnapGene Viewer	SnapGene	https://www.snapgene.com/
ApE plasmid editor	M. Wayne Davis, University of Utah	https://jorgensen.biology.utah.edu/wayned/ape/
Fiji (imageJ)	ImageJ	https://fiji.sc/
Genomic loci to compartment distance ImageJ plugin	Github	https://github.com/omidalam/compartment_dist
Python		https://www.python.org/
R		https://www.r-project.org/

STAR METHODS

• EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

a. CELL LINES

Mouse NIH 3T3 fibroblasts (ATCC CRL-1658) were grown in Dulbecco’s modified Eagle medium (DMEM with 4.5 g/l D-glucose, 4 mM L-glutamine, 1 mM sodium pyruvate and 3.7 g/l NaHCO₃) supplemented with 10% HyClone Bovine Calf Serum (Millipore-Sigma, Cat # 12133C). HCT116 and RPE1-hTert cells were obtained through the ATCC and cultured according to the 4DN recommended culture conditions (https://4dnucleome.org/cell-lines/). BJ-5ta-hTERT (ATCC CRL-4001), procured through ATCC, were cultured as per recommendations by ATCC (https://www.atcc.org/products/crl-4001).

• METHOD DETAILS

a. PLASMID DETAILS

Construction of retrofitting vector (pMini-HB-Tet96-attB-Neo-Puro-HSVTk-PB)

The plasmid pMini-HB-attB-Neo-Puro-HSVtk-PB (hereafter referred to as the retrofitting construct, Ctrl V1.0) was assembled in a stepwise manner. First, an intermediate plasmid, pMini-HB, was constructed by ligating an 826 bp PCR amplicon (5HB-BAC-fw / 3HB-BAC-rev) derived from the DHFR-RSLB1 BAC,³² containing sequences homologous to common BAC vectors (e.g. pBeloBac11 and pBACe3.6) into the pMiniT 2.0 vector (NEB, E1203S). A Gibson Assembly (NEB) approach was then used to insert three PCR-amplified fragments into pMini-HB: (i) a 2 kb attB–promoterless Kan/Neo cassette (attB-Neo_fwd / attB-Neo_rev) from the attB-Neo-SacB plasmid, (ii) a 382 bp 5′ piggyBac (PB) ITR fragment (5PB_fwd / 5PB_rev) from pJoyC030, and (iii) a 3.4 kb vector backbone fragment (pMini-HB_fwd / pMini-HB_rev) from pMini-HB.

The resulting construct, pMini-HB-attB-Neo-5′PB, was PCR amplified using 5HB_KpnI-fw / 5PB_NheI-rev to generate a 5.8 kb amplicon, which was digested with DpnI, KpnI, and NheI, and ligated to a 6.6 kb KpnI/SpeI fragment from the plasmid pHSV-JT96-Puro. This second fragment contained the 3′ PB ITR, attB-promoterless Kan/Neo, SV40 promoter-driven PuroR, and EF1α promoter–driven HSVtk counterselectable marker. The pHSV-JT96-Puro plasmid was previously generated by inserting a 1.4 kb SV40 promoter-PuroR cassette into the SpeI site of pHSV-JT96-EFS-attP-Bla. The attP sequence used in these constructs was obtained from the FRT-PGK-FRT-attP-Puro-DTA plasmid.

This retrofitting construct (Ctrl V1.0) was used for PB mobilization-based experiments. For attP-attB site-specific integrations, the retrofitting construct (Ctrl V1.0) was modified to a second-generation plasmid (pMini-HB-attB-Neo vector or Ctrl V2.0) through deletion of HSVtk-PB-Puro fragment. This deletion was achieved by digesting Ctrl V1.0 plasmid with SpeI and XbaI, followed by relegation of the 8.8 kb vector backbone. This modification of Ctrl V1.0 eliminates all the expressed eukaryotic cassettes, reducing the effect of integrated plasmid sequences on NS-targeting activity in our site-specific integration assays.

Retrofitting of BAC vector for stable cell line creation

The vector backbone of selected BACs was retrofitted with fragments shown in Figure 2F through λ Red-mediated BAC recombineering as per the published protocols.^32,33,36 A 10.0 kb retrofitting fragment with ~ 410 bp homology ends for recombineering was prepared from Ctrl V1.0 plasmid through XhoI and PacI restriction digestion. E. coli strain SW102 was used for BAC recombineering. Recombinants containing the retrofitting cassette were selected for resistance to kanamycin (100 μg/ml) at 30°C on LB agar plates. The integrity of BAC constructs was verified by restriction enzyme fingerprinting.

Construction of COL1A1 BAC derivative fragments

The construction of COL1A1 BAC fragments from COL1A1 BAC (RP11-267M22) involved several steps of BAC recombineering that enabled insertion or deletion at intended positions within the BAC. Initially, COL1A1 BAC intermediates F1, F2, and F3 were constructed by introducing the 3’PB-EF1α-BSD cassette from the template plasmid pJoyTet96lxpZeo-Bsd using primers specific to each derivative: 3PB-bsd-rec1-For and 3PB-bsd-rec1-newrev for F1, 3PB-bsd-rec2-For and 3PB-bsd-rec1-newrev for F2, and 3PB-bsd-rec3-For and 3PB-bsd-rec1-newrev for F3. Following this, the 5’PB-Tet96-Zeo cassette, derived from the same template by XbaI digestion, was subcloned into pMini-Col1A1-reg1 or pMini-Col1A1-reg2 at XbaI or NheI sites, respectively. The combinations for recombineering were as follows: COL1A1-F2 with pCol1A1reg1-5PB-Tet96 yielded COL1A1 F4 BAC; COL1A1-F3 with pCol1A1reg1-5PB-Tet96 yielded COL1A1 F5 BAC; COL1A1-F2 with pCol1A1reg2-5PB-Tet96 yielded COL1A1 F6 BAC; COL1A1-F3 with pCol1A1reg2-5PB-Tet96 yielded COL1A1 F7 BAC; COL1A1-F1 with pCol1A1reg1-5PB-Tet96 yielded COL1A1 F8 BAC; and COL1A1-F1 with pCol1A1reg2-5PB-Tet96 yielded COL1A1 F9 BAC. Further modifications through recombineering in BAC vector backbone were made to the F7 and F8 BACs by inserting a promoterless-Neo and attB fragment, resulting in COL1A1-F7attB and COL1A1-F8attB with kanamycin resistance.

The next phase involved Galk-based recombineering where F10attB was constructed by removing a 29.6 kb region from F7attB (Primers: recColD7GalkFor/ recColD7GalkRev; template pUGG plasmid; product size 1.3 kb). This F10attB BAC served as intermediate to create subsequent BACs F11 through F15. The primer combinations for GalK-based recombineering were as follows: COL1A1-F11 (recD11-D12fw/ recD11rev); COL1A1-F12 (recD11-D12fw/recD12-D15rev); COL1A1-F13 (recD13fw/ recD13-D14rev); COL1A1-F14 (recD14-15fw/recD13-D14rev); and COL1A1-F15 (recD14-15fw/recD12-D15rev). Galk-based 99 kb deletion in COL1A1 F8 BAC was made to derive COL1A1 F16 BAC (Primers recColD8GalkFor/ recColD8GalkRev; Template pUGG plasmid).

The intergenic construct COL1A1 F17 was made by ligating two PCR-amplified fragments derived from F13 BAC into the pMini2.0 vector. A 4.6 kb fragment from first intermediate construct (pMini-COL1A1 prod 1) was excised by BamHI and ligated to second intermediate construct (pMini-COL1A1 prod 2) at unique BamHI site. The final construct was verified by PCR and restriction digestion for correct orientation of insert. This 11 kb intergenic COL1A1-SGCA fragment was cut using PmeI/ZraI restriction enzymes and subcloned to PmeI digested retrofitting vector Ctrl V1.0 vector to get COL1A1-F17-RFTD construct.

COL1A1 F17 random deletion library

A random fragment library of a minimal speckle target region of COL1A1 BAC was prepared by excising and eluting a 11 kb fragment from COL1A1-F17-RFTD using PacI//PmeI/SpeI-HF (NEB). About 6 μg of the eluted linear plasmid was fragmented to 1000-9000 bp using a Bioruptor Pico (Diagenode) machine with a mode of 7 sec ON - 30 sec OFF for 1 sonication cycle. The sonicated DNA was eluted on agarose gel to retrieve fragments ranging from 2-8 kb, end repaired (End-It DNA End-Repair Kit: Epicentre Biotech ER0720), and blunt-end ligated to Ctrl V2.0 vector. The plasmid was digested with XbaI leaving behind a minimal vector containing-96mer, attP-Promoterless Kan/Neo cassette, followed by blunting by end-filling with Klenow (M0210, NEB) at 25°C. The ligated mix was transformed in NEB 5-alpha Competent E. coli (C2987H). Plasmids derived from randomly selected colonies were digested, and end sequenced to verify the inserts. The self-ligated vector without insert was used as the corresponding “empty vector” (Ctrl V2.0) negative control plasmid for experiments screening the fragment library.

Cloning of SPAD fragments

HCT116 template gDNA was used to PCR amplify fragments spanning H3K27ac peaks at chr12:6,531,445-6,536,855 (5411 bp), chr20:63,656,173-63,660,806 (4634 bp), and chr6:31,813,579-31,818,081 (4503 bp) using primers 369N23-D03-Fw/369N23-D03-Rev, 3184A7-D01-Fw/3184A7-D01-Rev, and D01_92G8_Fw/D01_92G8_Rev, respectively. The PCR products were gel purified and ligated to linearized Ctrl V1.0 vector.

Construction of TetR-TF fusion plasmids

Second generation lentiviral constructs expressing TetR-EGFP fused to transcription factors of interest were created by using Gibson Assembly approach (NEB). Briefly, the lenti F9-TetR-EGFP-IRES-Pur vector published elsewhere,⁷⁵ was linearized by BsrGI-HF restriction endonuclease (NEB), gel eluted and assembled with PCR amplified transcription factor CDS (coding sequence) as per manufacturers’ instructions. The CDS for TFAP2A, MXI1, ZBTB18, TWIST1, ZNF384 were amplified from cDNA prepared from U2OS cells (First strand synthesis kit, ThermoFisher Scientific), cloned in pMini2.0 vector (NEB), and verified by Sanger’s sequencing. The verified CDS were PCR amplified using indicated primers, incorporating a polyG-S (Glycine-Serine) linker to enable flexibility between TetR-EGFP and TF fusion protein. The codon-optimized ZNF460 CDS was synthesized by ThermoFisher, while CDS for TFAP2B (Cat no. DQ894621) and TFAP2C (Cat no. BC035664) were procured from Transomic technologies, Inc (Huntsville, AL, USA). The lentiviral constructs expressing TetR-EGFP fused AADs were generated by subcloning AAD fragments derived from full-length TFs mentioned above or AAD constructs described earlier.⁵¹ The AAD fragments were PCR amplified and subcloned between BsrGI/PshAI sites of lenti tetR-EGFP-IRES-Puro plasmid and verified by whole plasmid sequencing.

Mutational screen for COL1A1 D45 fragment

The predicted transcription factor binding sites on D45 fragment (628 bp) were eliminated in a synthetic fragment (F45-allmut), ordered through Integrated DNA Technologies (IDT). The product was PCR amplified (45frgE-fw/45frgA-Rev) and cloned into the PmeI/NdeI digested and dephosphorylated RFTDv2.0 vector (8.6 kb). The transcription factor binding sites in this construct were separately restored to the original individual or overlapping TF binding sites through site-directed mutagenesis. The primers used to create the final constructs (RFTDv2.0-D45-FragF rSDM1/2/3/4/5) are listed in the Table S1.

b. DRUG TREATMENTS

TRANSCRIPTIONAL INHIBITION

We applied two commonly used transcriptional inhibitors, Triptolide (TPL) and 5,6-Dichloro-1-β-D-ribofuranosylbenzimidazole (DRB). TPL irreversibly inhibits XPB/TFIIH at transcriptional initiation and induces rapid degradation of RNA Pol II,⁷⁶ while DRB targets the CDK9 kinase subunit of P-TEFβ and reversibly inhibits transcriptional elongation.⁷⁷ TPL treatment at 125 nM concentration for 1 hour itself is sufficient to induce global transcription block, however ~1% genes still retain paused Pol II on their promoters.⁷⁶ Therefore, to achieve maximum transcriptional inhibition in asynchronously proliferating hTERT-immortalized RPE1 we performed independent transcriptional inhibition experiments for extended periods (4 hours) with TPL or DRB. Transcriptional inhibition was performed for 4 hours on cells in log phase growth at 70% - 80% confluency with final concentrations of 0.1 μg/mL (278 nM) Triptolide, or 50 μg/mL (157 μM) DRB, or DMSO as the solvent-only control. In each experiment, we adjusted the volume of DMSO to make sure all groups of cells were exposed to the same DMSO concentration.

RPE1 cell synchronization was achieved by dual mitotic shake-off of cells. At ~80-90% confluency the asynchronous cells were subjected to a first mitotic shake-off to remove cells already in mitosis. The flasks were then rinsed with fresh media and treated with 50 ng/mL nocodazole in fresh media for 12 hours to arrest cells in mitosis. After 12 hours of incubation, we shook off the arrested mitotic cells (~4 million shaken off from one T300 flask) and transferred the cells within the nocodazole-containing media into new T150 flasks and added either 0.1 μg/mL Triptolide, DMSO, or nothing (for the “untreated” sample). The cells were incubated for an additional 1 hour for a combined mitotic and transcriptional arrest (the arrested cells won’t attach). After 1 hour of combined mitotic and transcriptional arrest, the cells were harvested by centrifugation at 200xg, and rinsed twice with fresh media containing triptolide (or DMSO or untreated control) to wash away nocodazole. The cells released from mitotic arrest into G1 phase were plated into fibronectin-coated T150 flasks and cultured in fresh media containing Triptolide (or DMSO) for 3 hours prior to cell harvesting. To coat flasks, fibronectin was diluted 1:100 in HBSS and added to flasks (15 mL for a T150 flask) and incubated at room temperature (RT) for 1 hour in cell culture hood immediately before use. After coating of the flasks, the fibronectin/HBSS was removed and cells/media were immediately added into the flasks.

EU-PULSE LABELING:

RPE1 cells were grown on glass coverslips with or without transcription inhibitors as described above. 5-Ethynyl uridine (50 mM stocks in 50% DMSO/50% PBS) was added to media at 1mM final concentration and pulse-labelled for 20 mins. After EU labeling the cells were rinsed with PBS three times, fixed with freshly made 3.7% paraformaldehyde at room temperature (RT) for 12 mins, washed with PBS for 3 x 5 mins, and then permeabilized with 0.5% triton X-100 /1X PBS (PBST) at RT for 15 mins. EU-labeled RNA was detected using Click-iT reaction cocktail containing Alexa 647-azide, according to the manufacturer’s instructions. Cells were incubated for 30 mins at room temperature (RT) in dark, followed by a wash with the supplied rinse buffer.⁷⁸

c. GENERATION OF LENTIVIRAL PARTICLES

Lentiviral particles were produced using the packaging plasmid pCMV-dR8.2 (Addgene plasmid #8455) and the envelope plasmid pCMV-VSV-G (Addgene plasmid #8454).⁷⁹ HEK293T cells were seeded in a 12-well plate and cultured to 60–80% confluency prior to transfection. For each lentiviral preparation, cells were cotransfected with 500 ng of the lentiviral vector, 450 ng of pCMV-dR8.2, and 50 ng of pCMV-VSV-G using lipofectamine 2000, following the manufacturer’s protocol. After 24 hours, the culture medium was replaced with fresh DMEM supplemented with 10% fetal bovine serum (FBS). Seventy-two hours post-transfection, viral supernatants were harvested and filtered through a 0.45 μm syringe filter to remove cell debris. All transductions were performed at a low multiplicity of infection (MOI). For transduction, 0.5-1 × 10⁵ cells were seeded per well in a 12-well plate and incubated with viral supernatant in the presence of 8 μg/mL polybrene for 24 hours. After 24 hours, the media was replaced with culture media supplemented with 1.0 μg/mL (NIH 3T3) or 1.2 μg/mL (HCT116) puromycin and maintained for another 4-5 days. For TetR-EGFP fusion protein tethering assays, NIH 3T3 cells were transduced with lentiviral particles expressing TetR–EGFP fusion proteins and selected with 1.0 μg/mL puromycin for 5 days prior to fixation.

d. GENERATION OF STABLE AND KNOCK-IN LINES

ESTABLISHMENT OF PB MOBILIZED CELL LINES:

BAC DNA for transfection of mammalian cells was prepared with the QIAGEN Large Construct Kit as per the manufacturer’s instructions. All other plasmid constructs for transfection were purified by Qiagen spin miniprep or midiprep kits. Lipofectamine 2000 was used to transfect the cells with BAC or plasmid DNA according to the manufacturer’s instructions. PB transposition of BACs was done by cotransfection of 10 μg BAC and 3 μg of inducible PB transposase expressing plasmid (mPB-L3-ERT2-tatRRR-mCherry). Nuclear translocation of the expressed PB transposase was achieved by supplementing the culture media with 2μM tamoxifen for 48 hours. The transfected cells were enriched in the appropriate culture medium supplemented by Puromycin for two weeks of puromycin selection, followed by 5 days of 10 μg/ml ganciclovir counterselection. This combination of positive and negative selection established a mixed clonal population of cells selected for BACs integrated via PB transposition rather than nonhomologous end joining (NHEJ). The ganciclovir-based negative selection eliminates cells containing BAC sequences integrated via NHEJ and thus still containing the HSVtk negative selection marker. However, the HSVtk negative-selectable marker located outside the PB ITRs, does not integrate in the host genome after PB transposition (Figure 2F, Figure S2D).

Mixed clonal populations, a polyclonal pool of independently derived stable transformants, were expanded and either transduced with lentiviruses or transfected with plasmids expressing TetR-EGFP for 4-5 days, as described in the preceding Section. Individual cell clones were obtained by serial dilution or retrieval of colonies using filter discs.⁸⁰

TARGETED KNOCK-IN OF NIH 3T3 CELLS:

NIH 3T3 cells were engineered to harbor a docking cassette comprising an attP site, a 96mer TetO array, and an EFS promoter-driven HSV-tk-IRES-BSD cassette at the indicated LAD or iLAD genomic loci (Figure S3). Genomic cleavage at the target sites – Reg 01 (Chr1:19564883–19568306, mm10) and Reg 03 (Chr1:105892227–105892428, mm10) was achieved by Cas9 and guide RNAs (gRNAs) expressed from the pX330A-1x3 vector (Addgene plasmid #58767).⁸¹ For each locus, two distinct gRNA constructs were designed. The gRNA oligonucleotides were cloned into the BbsI-digested pX330A-1x3 backbone via Golden Gate assembly as per the recommended protocol.⁸¹

Donor plasmids for homology-directed repair were constructed by amplifying ~2 kb genomic fragments- chosen to contain unique AgeI and SacII restriction sites towards the center of the fragments- from wild-type NIH 3T3 DNA using the following primer pairs: LAD1Fw/LAD1Rev (Reg 01) and LAD3Fw/LAD3Rev (Reg 03). Amplicons were cloned into the pMini2.0 vector (NEB PCR cloning kit). These plasmids were then used as templates to generate linear donor fragments via PCR, designed to remove gRNA target sites while retaining 800–900 bp homology arms on either side. Primer pairs used for generating linearized donor fragments were HB_LAD1_NFw/HB_LAD1_Rev and HB_LAD3_NFw/HB_LAD3_Nrev.

A donor cassette (~6.3 kb), including the attP-TetO-HSV-tk-IRES-BSD sequence, was PCR-amplified from the pDTA-J-EFS-attP-HSVtk-IRES-BSD-Tet96 plasmid using primers EFSattP_Fw and EFS_attP_Rev. The cassette was ligated into the homology arm-containing vectors via AgeI and SacII restriction sites to generate final donor constructs (LAD1-donor and LAD3-donor). These final donor constructs were linearized with NotI and PmeI, gel-purified, and co-transfected with their respective gRNA/Cas9 plasmids into NIH 3T3 cells at a 1:2 molar ratio (donor:gRNA plasmid).

SITE-SPECIFIC RECOMBINATION BY ΦC31 INTEGRASE:

Site-specific integration of specific DNA fragments at the intended locus within NIH 3T3 cells was achieved through the recombination of attB on BACs/plasmids and target attP sites previously integrated in the docking lines.⁷⁵ The φC31 integrase, derived from Streptomyces phage φC31, is a site-specific recombinase that enables precise, unidirectional integration of incoming construct through recombination between the bacteria-derived attB and phage-derived attP sequences.^82,83 This approach of integrating constructs through attP/attB-based site-specific integration is more suited than traditional CRISPR-based approach to integrate constructs larger than 5 kb.⁸⁴ The attB site on BACs/plasmids contain a promoterless Neo^R cassette positioned such that precise insertion at the attP docking site replaces the Blasticidin resistance cassette (BSD) from the docking target site with the Neo^R cassette, leading to expression of the neomycin resistance gene via the constitutive elongation factor 1 alpha (EF1α) promoter (Figure 3D).

To achieve integration of BACs/plasmids at either the iLAD (Reg 03) or LAD (Reg 01) docking sites, the BACs/plasmids and φC31 integrase expression vector plasmid (pCAG integrase) were cotransfected at a 1:2 molar ratio (BAC/plasmid:pCAG integrase). After 48 hours of transfection, the transfected NIH 3T3 cells were enriched in DMEM culture medium supplemented by G418 for three weeks, followed by 5 days of 10 μg/ml ganciclovir counterselection. Individual cell clones were obtained by retrieval of colonies using filter discs,⁸⁰ followed by transient expression of TetR-EGFP through lentiviral transduction for 4-5 days.

Genomic DNA for genotyping PCR from ~200,000 cells was isolated by overnight incubation at 60°C in a cell lysis buffer consisting of 10 mM Tris-Cl (pH 7.5), 10 mM NaCl, 10 mM EDTA (pH 8.0), 0.5% (w/v) Sarkosyl, and 1 mg/mL proteinase K. The isolated DNA was subsequently precipitated with salt and ethanol, followed by two washes in 70% ethanol. PCR amplification was performed using New England Biolabs (NEB) ThermoPol^® Taq DNA Polymerase to amplify target genomic regions. Specific primer sequences and expected amplicon sizes are shown in Figure S3E. The reaction mixture and cycling conditions followed the manufacturer’s recommended protocol for optimal target amplification. PCR products were subsequently resolved via agarose gel electrophoresis to determine the genotype of each sample.

e. IMMUNOFLUORESCENCE

Cells grown on glass coverslips were fixed in freshly prepared 2% paraformaldehyde (PFA) in 1XPBS buffer for 15 min at room temperature (RT). Cells were then washed with PBS for 2 x 5mins and with 0.1% PBST for 5 mins. Cells were blocked with 5% normal goat serum at RT for 1 hour and then incubated with rabbit anti-SON polyclonal antibody diluted 1:2000 and mouse anti-RNA Pol II Ser5P monoclonal antibody diluted 1:100 in blocking buffer at 4°C for 10-12 hours. Cells were washed with 0.1% PBST for 3 x 5 mins and then incubated with goat anti-rabbit-FITC and goat anti-mouse-Alexa594. Cells were washed with 0.1% PBST for 3 x 5 mins and then mounted with Mowiol-DABCO anti-fade media containing 0.3 μg/ml DAPI.⁸⁵

f. FLUORESCENCE IN SITU HYBRIDIZATION (FISH)

DNA FISH probes:

GAPDH, COL1A1, Dhfr/Msh3, and COL1A2 BACs were used to prepare probes for 3D DNA FISH. A biotinylated or Cy3 conjugated oligonucleotides (see Table S1) synthesized at IDT were used as probe for FISH based detection of the TetO array. Preparation of biotin or digoxigenin labeled DNA FISH probes and 3D DNA FISH of interphase nuclei were carried out as described earlier,¹⁸ with small modifications. Instead of culturing cells on poly-L-lysine coated coverslips for 15 minutes, the cells were grown on untreated glass coverslips (12 mm diameter) for 3-4 days before fixation with 2% paraformaldehyde in 1XPBS for 15 mins at RT. FISH signals were detected by incubation with Alexa Fluor 647 conjugated Streptavidin (1:200 dilution) or Alexa 594 conjugated Streptavidin (1:200 dilution) for biotin-labeled probes, or Alexa Fluor 647 conjugated IgG fraction monoclonal mouse anti-digoxin (1:200 dilution) for digoxigenin labeled probes, diluted in SSCT with 4% Bovine Serum Albumin, for 2 hours at RT or overnight at 4°C. Coverslips were washed in SSCT for 4 × 5 min, rinsed with 4x SSC and mounted. All samples were mounted with a Mowiol-DABCO anti-fade medium containing ~0.3 μg/ml DAPI.⁸⁵

TetO visualization in the tethering assay was performed using the RASER-FISH method.⁸⁶ RASER (resolution after single-strand exonuclease resection) FISH uses exonuclease digestion of newly synthesized DNA strands to allow probe hybridization without heat denaturation or acid treatment, thereby better preserving nuclear morphology. Briefly, cells grown on 12-mm diameter coverslips were labelled with a BrdU/BrdC mix (3:1) at a final concentration of 10 μM for 15 hours. Cells were fixed in 2% paraformaldehyde in 1X PBS for 15 min at RT, washed three times in 0.5% PBST for 5 min each, and rinsed two times with 1X PBS. Cells were then stained with Hoechst 33258 (0.5 μg/mL in 1XPBS), exposed to 365 nm wavelength UV light for 30 min, and treated with Exonuclease III (NEB; 5 U/μL final concentration) at 37°C for 30 min in humidified chamber. Labelled TetO probes (2 ng) and BAC probes (200 ng) in 1X hybridization buffer were preannealed 37°C for 30 min, followed by denaturation at 75°C for 3 min. Coverslips were hybridized with probes for 72 hr at 37°C in humidified chamber. Following hybridization, coverslips were washed two times with prewarmed 2XSSC and once with 1XSSC at 37°C for 30 min each. FISH signals were detected using Alexa Fluor 647-conjugated streptavidin or Alexa Fluor 594-conjugated streptavidin (1:200 dilution for biotin-labeled probes) diluted in SSCT containing 4% bovine serum albumin. Detection was carried out for 2 hours at RT or overnight at 4°C. Coverslips were washed four times in SSCT for 5 min each, rinsed once in 4XSSC, and mounted in Mowiol-DABCO antifade medium containing.⁸⁵

g. TSA-SEQ

SON TSA-Seq 2.0 procedure with enhancement condition E, reagents used, and data analysis protocols has been described previously.²⁰ Specific conditions used for these data sets, such as antibody concentrations and source, numbers of cells, library construction, etc, are provided in the metadata for each dataset. The datasets and corresponding metadata sheets are publicly available at the 4D Nucleome server with accession numbers provided in the Key resources table.

h. ChIP-qPCR

Chromatin immunoprecipitation (ChIP) was performed as described previously,⁸⁷ with minor modifications. Briefly, approximately 10 million NIH 3T3 cells with WT F45 or all-mut F45 fragment integrated at iLAD locus were used per experiment. The cells cultured in T300 flasks were crosslinked with 1% formaldehyde in 1X PBS for 10 min at room temperature, and the reaction was quenched with 200 mM glycine for 5 min. Cells were washed with ice-cold PBS and lysed in 1X RIPA buffer (Cell Signaling Technology) supplemented with protease inhibitors (Roche).

Chromatin was fragmented to an average size of 200-500 bp by sonication using a Bioruptor Pico (Diagenode; 40 sec ON - 30 sec OFF, 20 cycles). Following centrifugation, an aliquot of chromatin was reserved as input control and the remaining chromatin was incubated overnight at 4°C with 2 μg of anti-H3K27ac, anti-ZNF384, or normal rabbit IgG antibodies pre-coupled to Protein G magnetic beads (Cell Signaling Technology).

Bead-bound chromatin complexes were washed five times with ChIP Wash Buffer (Santa Cruz Biotechnology) and once with TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA). Chromatin was eluted in 1% SDS in TE buffer at 55°C for 30 min and reverse crosslinked overnight at 65°C. Samples were subsequently treated with RNase A (Qiagen) for 30 min at 37°C and then Proteinase K (New England Biolabs) at 55°C, followed by DNA purification.¹⁸ ChIP enrichment was quantified by qPCR using two independent primer pairs for each genomic region analyzed. Ct values from the two primer pairs were averaged and enrichment was first calculated as percentage of input chromatin (% Input). The resulting % Input values for H3K27ac or ZNF384 ChIP samples were then normalized to the corresponding normal rabbit IgG control, which was set to 1. Data are presented as fold enrichment relative to IgG for two independent replicates.

• QUANTIFICATION AND STATISTICAL ANALYSIS

a. IMAGE ACQUISITION AND ANALYSIS

All images were acquired as 3D optical-section images with 0.2 μm z-steps using a V4 OMX (GE healthcare) microscope equipped with a 100X, 1.4 NA oil immersion objective (Olympus) and two Evolve EMCCDs (Photometrics). Image stacks were subjected to 10 cycles of constrained iterative deconvolution (conservative Ratio method) to remove out of focus blur, and corrected for camera and chromatic aberration misalignments using OMX image alignment by local triangulation method using SoftWoRx software (version 7.0) (GE Healthcare). All subsequent image analysis and preparation were done using Fiji software (ImageJ). Representative images were assembled using Illustrator (Adobe).

Sum intensity measurement for Figure 1 was made in FIJI on non-deconvolved images where the total intensity of 26 slices (0.2 μm in z-axis) was normalized to cytoplasmic background, exposure time, and % transmitted exciting light. The quantification methods were adapted from a previous publication.⁸⁸

The SON antibody used in our study stains the core of NS,^9,15 while the periphery of NS composed of U1, U2 snRNPs extends to an additional 20-50 nm beyond the NS core.⁹ Earlier the gene was considered NS associated when the gene was in physical contact with NS as observed with the fluorescence microscopy.^17,89 However, accommodating the resolution of limits of fluorescence imaging we used a cut-off for NS-association when the transgene signal was detected within 0.25 μm of NS periphery. To ensure reproducibility of measurements and to negate the effects of background staining as well as the variable intensities of different NS, we defined the edge of the NS by where the pixel intensity fell to 40% of the maximum SON staining intensity of that NS. Distances relative to the nearest edge of the nearest NS were measured by using the Loci-Compartment Plugin for imageJ. The plugin is freely available and can be accessed at https://github.com/omidalam/compartment_dist.

b. STATISTICAL ANALYSIS AND REPRESENTATION

Violin plots within figures show the data distribution of individual measurements for the indicated treatment or tested constructs. Each violin represents the kernel density of individual measurements, with embedded boxplots showing the mean (solid black squares), median (black horizontal lines within box), 25% and 75% percentiles (box limits) and range within 1.5x IQR (top and bottom whiskers around box). Individual data points are overlaid as jittered dots (gray diamonds). Treatment reagent or construct are indicated on the x axis, while the y axis denotes the measurement distribution and individual data points.

A z-test for two population proportions was used to assess statistical differences in the proportion of events within the 0–0.25 μm bin of the stacked histograms. The test was performed under the null hypothesis that the proportions in the two groups are equal, and corresponding p-values were calculated accordingly. The null hypothesis was tested using this equation:

$$z=\frac{(p_1-p_2)-0}{\sqrt{p(1-p)(\frac{1}{n_1}+\frac{1}{n_2})}}$$

Here p₁ and p₂ are the sample proportions from the two groups, n1 and n2 are the sample sizes of the two groups, and p is the pooled proportion calculated as:

$$p=\frac{x_1+x_2}{n_1+n_2}$$

x1 and x2 are total counts in the bin (0-0.25 μm) from the two groups.

Statistical significance of differences in mean distances between samples was assessed using one-way ANOVA followed by a post hoc Tukey’s Honestly Significant Difference (HSD) test. P-values are represented in the figures as follows: n.s., not significant; *p < 0.05; **p < 0.01; ***p < 0.001 ; **** p < 0.0001. Unless otherwise specified, data are presented as mean ± SEM (standard error of the mean). Statistical analyses were performed using Microsoft Excel, the R package ‘stats’ (for ANOVA and HSD tests), and the online tool at https://www.socscistatistics.com.

c. DATA DESCRIPTION

The TSA-Seq data generated in this study have been submitted to the 4DN Data Portal (https://data.4dnucleome.org/). The corresponding accession numbers can be found in the key resources table.

Supplementary Material

1

Figure S1. Genome organization relative to NS is maintained independent of transcription in log phase HCT116 cells, related to Figure 1.

(A) Genome browser view of SON TSA-Seq enrichment in untreated, DMSO-, or TPL-treated cells.

(B-C) 2D histograms comparing SON TSA-Seq scores for DMSO replicates (B) or DMSO versus TPL (C). Colors show 20 kb bin counts (~0.02 × 0.02), Pearson’s r correlation.

(D) Representative image panels and cartoon illustrating NS organization during interphase, mitosis, and early G1 phase in RPE1 cells. Scale bars= 5 μm.

Figure S2. Certain endogenous genomic loci associate with NS in transcription-independent manner in HCT116 cells, related to Figure 2.

(A) Genome browser view of GAPDH, COL1A1, and COL1A2 loci showing proximity to NS by SON TSA-Seq (top tracks) and expression levels by RNA-seq (bottom tracks).

(B-C) Visualization and quantification of distance distribution of endogenous GAPDH, COL1A1, and COL1A2 genomic loci (3D DNA FISH, red) relative to NS (green) and DAPI (blue) after DMSO or DRB treatment (4 hours). Images in each panel are at same magnification, scale bars= 2 μm (insets, 0.5 μm) in indicated human cell lines (n = 162-169). DRB treatment compared with DMSO controls.

(D) Schematic of PiggyBac (PB) transposition strategy used for candidate BAC screening in HCT116 cells.

Figure S3. Generating and validating NIH 3T3 docking lines for φC31-mediated site-specific integration, related to Figure 3.

(A) Chromosome-wide Lamin B1 and SON TSA-Seq profiles in mouse cells highlighting regions (red) selected for CRISPR-Cas9 insertion of an attP site, TetO-96mer array, and EFS-HSVtK-IRES-BSD cassette.

(B) Maximum-intensity projections of two optical z-sections from representative nuclei showing TetO arrays (green) integrated at indicated loci relative to NS (SON, red), and DAPI (blue).

(C-D) Quantification of TetO-NS distance distribution in indicated NIH 3T3 knock-in clones as stacked histograms (C) and as violin plots (D). (n=87-99).

(E) Schematic representation of clone validation by genotyping PCR. Primer combinations and expected product size are shown for 5’ and 3’-junctions.

(F) Maximum-intensity projections of three optical z-sections from representative nuclei of iLAD clone g1-A2. White squares mark TetO FISH (green) overlapping the true insertion site visualized by BAC FISH (red). Multiple BAC FISH signals indicate NIH 3T3 chromosome 1 polyploidy. DAPI (blue).

Images in each panel are at same magnification, scale bars= 5 μm (insets, 0.5 μm).

Figure S4. Predicted transcription factor binding motifs within the 600 bp COL1A1 fragment F45, related to Figure 4. Sequence logos of TF binding motifs predicted using human JASPAR 2022 database. The x-axis shows position within the consensus motif and the y-axis shows the information content (bits), reflecting the degree of nucleotide conservation. Nucleotide height reflects frequency and contribution to binding specificity. Below each motif, the corresponding 600 bp WT sequence and mutant versions are shown; mutated bases are highlighted (red) and altered motif shaded (gray).

Figure S5. Predicted acidic activation domains (AADs) in the indicated transcription factors identified with ADpred, related to Figure 5.

(A) Predicted AADs in the indicated TFs generated using ADpred; x-axis, amino acid position within the protein; y-axis activation domain probability. Red line indicates prediction threshold. Colored plots indicate AADs (blue) or non-acidic activation domains (NAADs) or CTCF protein (red). Peptides tested in the tethering assays (shaded gray).

(B) Summary of peptide sequences tested in tethering assay, including protein name, UniProt ID (canonical isoform), peptide sequence, and position. Acidic residues (D, E) are green, hydrophobic residues (F, M, I, L, V, C, W) red, and basic residues (K, R) blue.

2

Table S1. Plasmid/BAC constructs and oligonucleotides used in the study, related to the STAR methods.

3

Table S2. All datasets and statistical analyses used to generate plots in main and supplementary figures, related to the STAR methods.

HIGHLIGHTS.

• Nuclear speckle-genome association occurs without active transcription

• Speckle-associated DNA sequences autonomously target transgenes to speckles

• Nuclear speckle targeting is mediated by multiple redundant cis elements

• Nuclear speckle targeting is a common activity of acidic activation domains

Data And Code Availability

• Sequence data generated in this paper are available at the 4D Nucleome Data Portal (https://data.4dnucleome.org), with accession numbers as listed in the Key Resources Table. Source data for Figure images is available in Mendeley Data, V1, doi: 10.17632/rjwkxm566p.1. All data used for plots in Figures and Supplementary Figures is supplied in Supplementary Data as Table S2.

• The Loci-Compartment Plugin for imageJ was used to measure distances relative to NS. The plugin is freely available and can be accessed at https://github.com/omidalam/compartment_dist.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.